Ivy League Failures
Back in October of 2014, Larry Milsten of the Yale Daily News reported on a data server failure that brought down an entire university’s website and email accounts, during the wonderfully inopportune time of midterms.
That University of course, was Yale. In that October 9th article, Milsten interviewed multiple sources directly involved in Yale’s ITS division, from students and library workers to the current University CIO, Len Peters.
The article walks through the different ways a data server shutdown influenced the campus, while also trying to find the root cause of the issue. In the article, Peters describes how rare the occasion was for Yale, as the server experienced a multi-point power failure. The data center failure left students without their university emails, without access to online courses; the failure even took down the Yale library.
The piece of Milsten’s article that sticks out the most to us was the "Why?" of it all. Specifically, two instances in which a librarian and a student working for Yale ITS each named "overheating" as the reason for the data center issues. Both Rowillie Ross and Lizza Rodler believed that the failure was because of an air conditioning mishap.
The October 9th article was not the end of the investigation. After the weekend, on October 13, 2014, Milsten reported that Yale ITS was still investigating the Thursday server failure. While the article eventually divulges into asking Yale students how they dealt with the power outage, the opening line of the article sticks out:
Three days after the largest computer system failure in recent memory at Yale, the root cause has yet to be fully determined.
To make matters worse . . . the Yale servers crashed again only a few weeks later.
The data server failing for a few hours at Yale is one thing. What happens when a data center, serving more than a University, fails and all its data is lost?
While it was simply word-of-mouth reporting from those two individuals, it did peak our interest back in 2014, as temperatures adverse influence on data center security is a problem that we've stumbled across before. While the cause of Yale's server failures was found to be to to electrical and power failures, the question remains: can server rooms overheat? If so, is it common? How big of a problem is it?
It's a big one. Servers room with racks upon racks of data storage and humming servers generate a lot of heat. And they have to be kept cool, because overheating leads to disaster for data.
The world’s data is stored in server rooms across the country. There really is no "magic cloud." The cloud is a data center or server room, that allows you to access your Facebook, Gmail, and Bank of America data by logging onto a website. Small businesses and large corporations have their data in server rooms. I know we do here at Dickson. For all the media coverage of hackers and viruses, one of the scariest threats to our data is hardly talked about: temperature.
Because server and data centers use so much power, if the HVAC or cooling system in a data center fails, the temperature of those rooms rises very quickly. Many times, those failures are the result of a power failure. When the power goes out, data centers will continue to run on back-up generators. While the HVAC and Cooling Systems (which sometimes take up as much if not more power than the data centers themselves) cease operation. Because server and data centers use so much power, they generate a lot of heat. So much so, that a simple Google Search for "server room temperature" returns results like: "Germans get free heating from the cloud" and "Heating a skyscraper with a data center."
As temperatures rise, computers fry. This leads to extremely high energy costs for data centers, so much so that there has been a call from tech giants like Google to "Raise Your Data Center Heat." All this contention and information comes before the many data-center hacks that outline the solution to your heating problem hinges on the exact number of perforated tiles your data needs per square foot of space.
It can all be a bit overwhelming. While we will leave the control of your data center’s temperature up to you, we can offer some advice in the monitoring department.
Monitoring the temperature of your data center gives you a few tools that can be essential to the security of the data you are storing:
Alarms are pretty straight forward. When the temperature of your data center gets too high (or much less commonly, too low) data loggers and data monitoring systems can alert you via text, email, or a phone call.
A data logger can be your best friend and saving grace.
At Dickson, we offer devices that use cloud storage, to keep your cloud safe! Specifically, our line of DicksonOne temperature monitoring data loggers can alert you when your server room gets too hot: via text, email, or phone call.
Mapping? That’s a little more nuanced. We sell data loggers to a lot of warehouses, and those warehouses place loggers across their entire floor plan to account for temperature differences and stratification that they may not be accounting for with their current monitoring system. Mapping is also extremely useful for data centers. Data centers have racking just like warehouses. So, they also have temperature stratification, a concern with air flow, and problem spots. Using data loggers (like these!) to map your data center floor will help you find those problem spots, and also help you find inefficiencies in your cooling system.
However you handle the temperature of your data center, do it completely and thoroughly. We believe part of that is using a data logger to monitor the temperature of your servers.